A big reason that token costs become an issue in organizations is leadership is not making decisions about how to use AI, the processes that need to change as a result, & how to think about multiplayer use. Solving it by rationing intelligence first is a pretty blunt strategy
ECONOMY
-
Sierra discusses outcome pricing versus usage pricing for AI agents
By
–
Selling a membership or closing a car sale? Outcome-based pricing makes sense. The value is real and measurable.
— LangChain (@LangChain) 29 juin 2026
Answering a knowledge lookup? Usage-based is simpler and probably right.@zackrw from @SierraPlatform on how AI pricing should actually map to value delivered. pic.twitter.com/XveYoWBP6eSelling a membership or closing a car sale? Outcome-based pricing makes sense. The value is real and measurable. Answering a knowledge lookup? Usage-based is simpler and probably right. @zackrw from @SierraPlatform on how AI pricing should actually map to value delivered.
-
Commentary compares local model hardware costs with cloud subscriptions
By
–
So you’re break-even after around 2.5 years compared to a sub. Flash compares to Sonnet/Terra which is close to unlimited use. They are cool devices tho, and being able to run everything locally is great. Not an economical case but neat for research.
-

Commentary proposes return on tokens as a measure of AI value
By
–
ROT — Return on Tokens. We all knew we would end up here at some point. Tokenmaxxing was a dumb idea from the start. It’s just that when everyone is drinking from the coolaid it’s hard to resist. But for some of us it’s been clear as day for years that first we will see
-
Commentary links GLM progress to model convergence and pricing pressure
By
–
GLM 5.2 is the ultimate culmination of what i have argued here for the last three years. no technical moat and widely known formula -) convergence -) price wars -) small or negative margins.
-
Commentary discusses frontier pricing pressure and generative AI margins
By
–
Thanks for engaging.
1. I expect price wars at the frontier soon (and one could argue there has already been some strong downward pressure).
2. Curious who in the GenAI world aside from Nvidia has solid margins. (Suno is one example but there don’t seem to be many.)
3. My claims -
Commentary questions Fable delay’s impact on Anthropic’s June revenue
By
–
do you want to bet with me on the revenue of Anthropic in June? Bet it was massive and didn't slow down one bit because of the delay of release of Fable 5.
-
AI data center growth drives demand for gas turbines
By
–
How The AI Boom Is Causing A Surge In Demand For Gas Turbines? https://
youtu.be/gk8ziQCWK1w?si
=oL1_OWAo_tiW5Opj
… via @YouTube #AI #artificialintelligence #datacentre #gasturbine #tech @AlbertoEMachado @Eli_Krumova @postoff25 @Khulood_Almani @anand_narang @NutritiousMind @baski_LA @TanyaSinha_ -
OpenAI fails to provide GDPval measure for GPT 5.6
By
–
Annoying that OpenAI doesn’t seem to give a GDPval measure for GPT 5.6. One of the best measures of economically valuable work.
-
Open models have better dollar per token than closed APIs
By
–
An interesting way to take Noam at his word in regards to always keeping a constant inference budget for any eval reporting –
— swyx @aiDotEngineer WF (@swyx) 27 juin 2026
is that open models have a lot more dollar per token mileage than closed model APIs. So anyone launching an open model today or situationally… https://t.co/vGl7tQZgWnAn interesting way to take Noam at his word in regards to always keeping a constant inference budget for any eval reporting – is that open models have a lot more dollar per token mileage than closed model APIs. So anyone launching an open model today or situationally